Lessons Learnt and Future Improvements

The results were initially disappointing, and it is clear from these trials that the procedure of attaching ISFET sensors to underwater gliders still requires some development. However, by correcting the pH measurements, the potential usefulness of deploying ISFET sensors was illustrated.

The stability of the glider-integrated and stand-alone p(CO2) ISFET sensors was poor. It is

unclear why there was a problem with these measurements. However, the stability of the sensors may in part be related to the gas-permeable membrane. To measure p(CO2), the pH

of the water is measured within a gas permeable membrane filled with an inner solution. Variations in the pH of this inner solution is related to the penetration of CO2through the

membrane, and is used to derive p(CO2). It may have been the case that the diffusion

of CO2 across or within the membrane was highly variable. Furthermore, relating to the

glider-integrated dual ISFET pH - p(CO2) sensor, it was thought that there was an issue

with the electricity supply from the glider. It was thought that the regular on/ off cycling of power to the integrated sensor in between sampling did not allow it to function properly. The addition of backup batteries to supply the integrated sensor in between sampling was tested during the BOUSSOLE deployment. This worked for the glider-integrated ISFET pH sensor. However, the glider-integrated ISFET p(CO2) sensor still encountered stability

issues. The cause of this is unclear.

Due to thesis time constraints and the poorer quality p(CO2) measurements, the method of

converting raw ISFET p(CO2) counts (corresponding to the inner-membrane solution pH)

to quantitative p(CO2) units was not explored. In previous work, raw ISFET p(CO2) counts

were used to provide qualitative discussion only (e.g. Shitashima et al. (2013, 2008)). In future, a method using reference solutions of known p(CO2) to calibrate p(CO2) counts

before and after the deployment could be developed.

The stability of the stand-alone ISFET pH sensor was poor. The multi-directional drift was inconsistent over time and spatially-independent. It was clear that this multi-directional drift was related to an issue with the ISFET sensor, rather than the environmental conditions at the time. It is thought that a change in the interface potential (E∗) between the two n type silicon parts of the semiconductor might be responsible for this drift. To elucidate this drift further, in future two ISFET sensors should be tested in laboratory conditions within a bridge circuit to attempt to isolate possible factors contributing to drift. Focussing on the root cause of the sensor drift, rather than correcting the pH data for drift after the deployment, would be more beneficial to the longterm study of ISFET pH-p(CO2) sensors.

It is also possible that the limited pre-conditioning of the sensor in local seawater caused the sensor to drift during the deployment. Pre-conditioning an ISFET sensor prior to a deployment limits the effect of bromide on sensor performance, and an asymptotic drift can be experienced when the sensor is in contact with water. Such drifts usually last a couple

of days. The ISFET stand-alone pH sensor was preconditioned for one hour only due to time constraints, although pre-conditioning is recommended for some weeks. In future, it is recommended to pre-condition the sensor for at least one week prior to a deployment.

The sensor was apparently affected by temperature and pressure during the REP14 and

BOUSSOLE deployments. It is unclear to what extent the empirical relationship between insitu temperature, pressure and pH can be generalised. To explore these relationships further, it is recommended to follow similar steps to those described by Johnson et al. (2016). Their ISFET DuraFET sensor was tested for its response to temperature and pressure changes using a custom-built titanium chamber. The sensor was calibrated by cycling through the typical range of temperature and pressure that may be experienced by the sensor when in the field.

The effect of light on the ISFET pH - p(CO2) sensor was unexpected. This effect had not

been highlighted in previous work (e.g. Shitashima et al. (2013, 2008)), as the focus had

been on observing deep measurements (e.g. monitoring seabed CO2 leakages) on shorter

timescales. It was suggested for the BOUSSOLE trial to either surround the sensors with a light shield or to place the sensors on the underside of the glider. Placing the sensor on the underside of the glider during the BOUSSOLE deployment significantly limited the affect of light on pH measurements.

The development of the ISFET pH - p(CO2) sensor during the course of this study was

limited due to practical issues. The sensors were loaned to the University of East Anglia for the duration of the deployments, and for a period of two days in the laboratory. This meant that it was not possible to explore the issues raised in this work further over a longer time period in the laboratory at the University of East Anglia. Due to this limitation, it was also not possible to pre-condition the sensors for a longer time period prior to the deployments.

At the current stage of development, the ISFET pH - p(CO2) sensor can be used on

autonomous platforms to measure pH in the Mediterranean Sea and globally. However, time should be invested to properly pre-condition the sensor before deployment, and to collect reference ship pH samples to deal with likely instrumental issues. Although the calibration equations derived using the relationship between pH and temperature, and pH and pressure in the northwestern Mediterranean Sea worked in reducing the large range of variability associated with instrumental issues, it is unclear whether they could be applied

to pH measurements obtained in other regions. However, considering that each trial

occurred at a different location, a different time of the year, and in vastly different environments (i.e. mixing vs. stratification), the similar slope coefficients suggest that the derived ISFET pH sensor corrections are applicable to a wide range of situations. Unless improvements are made to the experimental ISFET pH - p(CO2) sensor used in this study,

it is suggested that those wishing to measure insitu pH autonomously with high spatial and temporal resolution as part of an established observing program explore other ISFET